MINGLE

An important but under-explored problem in computer science is the automated analysis of conversational dynamics in large unstructured social gatherings, such as networking or mingling events. Research has shown that attending such events contributes greatly to career and personal success.

While much progress has been made in the analysis of small pre-arranged conversations, scaling this up robustly presents a number of fundamentally different challenges. Unlike analysing small pre-arranged conversations, sensor data is seriously contaminated during mingling: audio by background chatter; video by people occluding each other; and proximity by noisy radio reflections given the high density of human bodies. Moreover, determining who is talking with whom is difficult because groups can split and merge at will.

A fundamentally different approach is needed to handle both the complexity of the social situation and the uncertainty of sensor data, which has not been addressed by state-of-the-art techniques. By exploiting people's non-verbal behaviour, we will develop novel learning methods to estimate the quality of conversational interactions in mingling events. A fresh perspective is needed to solve this, by relying more heavily on sensors that can capture body motion. The motivation for this comes from findings in social and behavioural psychology demonstrating links between body movements and conversational events.

Novel representations of jointly coordinated behaviour will be developed to detect conversational events using multi-sensor streams. By departing from current graph-based representations of groups, MINGLE proposes to both predict and detect the evolution of conversation partners by combining each individual's intrinsic motivations with the emergent behaviour of their conversational group. Solving this problem will provide the breakthrough necessary to advance significantly a number of other domains, from human-robot interaction to organisational psychology.